关键词: 卷烟, 配方模块, 价值评价, 历史生产数据, BP神经网络算法

Abstract: In response to the shortcomings of subjectivity and blindness in traditional formula module quality evaluation methods, as well as the significant deviation between the quality of the formula module and its actual production and use,this study proposed a value calculation scheme for the formula module based on historical production data, mining the actual usage of the formula module and measuring the relationship between the value and quality of the formula module. The calculation results of this scheme were used as indicators to characterize the comprehensive quality of the formula module, and the rationality of the calculation results was verified using BP neural network algorithm and statistical analysis methods. The results showed that sensory indicators （aroma, smoke, taste, aroma type） were significantly positively correlated with grade, while chemical indicators （nicotine CV, sugar to alkali ratio） were not significantly correlated with grade. Grade as an evaluation scheme for the quality of the formula module was too subjective. The value of the formula module was significantly correlated with sensory and chemical indicators, indicating that the value of the formula module combined subjective and objective indicators, making it more suitable to represent the quality of the formula module;the value calculation scheme of the formula module fully considered the characteristics of actual production and operation of the cigarette, which was more in line with production reality and could provide better guidance for the production and use of the formula module of the cigarette.

Key words: cigarette, formula module, value evaluation, historical production data, BP neural network algorithm